Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

OpenEM: Large-scale multi-structural 3D datasets for electromagnetic methods

Shuang Wang, Xuben Wang, Fei Deng, Peifan Jiang, Jian Chen, Gianluca Fiandaca

TL;DR

OpenEM tackles the shortage of large-scale 3D geoelectric datasets by providing a unified, multi-structural OpenEM model suite with ~1.08 million 3D resistivity models (3–7 layers, 1–5 anomalies) spanning nine geological types. Models are generated using von Kármán covariance priors and iterative faults/folds, with anomalies embedded to reflect geological realism. To enable DL-assisted EM tasks, a 3D U-Net–based forward model with altitude embedding delivers rapid, reliable predictions for AeroTEM IV, achieving sub-5% error on most cases and under 1% in simpler scenarios. The dataset and open-source forward modeling tools are released to accelerate deep learning applications in electromagnetic methods and to support uncertainty quantification and generalization studies across geologies and survey systems.

Abstract

With the remarkable success of deep learning, applying such techniques to EM methods has emerged as a promising research direction to overcome the limitations of conventional approaches. The effectiveness of deep learning methods depends heavily on the quality of datasets, which directly influences model performance and generalization ability. Existing application studies often construct datasets from random one-dimensional or structurally simple three-dimensional models, which fail to represent the complexity of real geological environments. Furthermore, the absence of standardized, publicly available three-dimensional geoelectric datasets continues to hinder progress in deep learning based EM exploration. To address these limitations, we present OpenEM, a large scale, multi structural three dimensional geoelectric dataset that encompasses a broad range of geologically plausible subsurface structures. OpenEM consists of nine categories of geoelectric models, spanning from simple configurations with anomalous bodies in half space to more complex structures such as flat layers, folded layers, flat faults, curved faults, and their corresponding variants with anomalous bodies. Since three-dimensional forward modeling in electromagnetics is extremely time-consuming, we further developed a deep learning based fast forward modeling approach for OpenEM, enabling efficient and reliable forward modeling across the entire dataset. This capability allows OpenEM to be rapidly deployed for a wide range of tasks. OpenEM provides a unified, comprehensive, and large-scale dataset for common EM exploration systems to accelerate the application of deep learning in electromagnetic methods. The complete dataset, along with the forward modeling codes and trained models, is publicly available at https://doi.org/10.5281/zenodo.17141981.

OpenEM: Large-scale multi-structural 3D datasets for electromagnetic methods

TL;DR

OpenEM tackles the shortage of large-scale 3D geoelectric datasets by providing a unified, multi-structural OpenEM model suite with ~1.08 million 3D resistivity models (3–7 layers, 1–5 anomalies) spanning nine geological types. Models are generated using von Kármán covariance priors and iterative faults/folds, with anomalies embedded to reflect geological realism. To enable DL-assisted EM tasks, a 3D U-Net–based forward model with altitude embedding delivers rapid, reliable predictions for AeroTEM IV, achieving sub-5% error on most cases and under 1% in simpler scenarios. The dataset and open-source forward modeling tools are released to accelerate deep learning applications in electromagnetic methods and to support uncertainty quantification and generalization studies across geologies and survey systems.

Abstract

With the remarkable success of deep learning, applying such techniques to EM methods has emerged as a promising research direction to overcome the limitations of conventional approaches. The effectiveness of deep learning methods depends heavily on the quality of datasets, which directly influences model performance and generalization ability. Existing application studies often construct datasets from random one-dimensional or structurally simple three-dimensional models, which fail to represent the complexity of real geological environments. Furthermore, the absence of standardized, publicly available three-dimensional geoelectric datasets continues to hinder progress in deep learning based EM exploration. To address these limitations, we present OpenEM, a large scale, multi structural three dimensional geoelectric dataset that encompasses a broad range of geologically plausible subsurface structures. OpenEM consists of nine categories of geoelectric models, spanning from simple configurations with anomalous bodies in half space to more complex structures such as flat layers, folded layers, flat faults, curved faults, and their corresponding variants with anomalous bodies. Since three-dimensional forward modeling in electromagnetics is extremely time-consuming, we further developed a deep learning based fast forward modeling approach for OpenEM, enabling efficient and reliable forward modeling across the entire dataset. This capability allows OpenEM to be rapidly deployed for a wide range of tasks. OpenEM provides a unified, comprehensive, and large-scale dataset for common EM exploration systems to accelerate the application of deep learning in electromagnetic methods. The complete dataset, along with the forward modeling codes and trained models, is publicly available at https://doi.org/10.5281/zenodo.17141981.
Paper Structure (10 sections, 4 equations, 8 figures, 3 tables)

This paper contains 10 sections, 4 equations, 8 figures, 3 tables.

Table of Contents

  1. Methodology
  2. Half-space models with anomalies
  3. Multi-structure models
  4. OpenEM
  5. Forward
  6. Network architecture
  7. Train
  8. Test
  9. Discussion
  10. Code and data availability

Figures (8)

  • Figure 1: Examples of half-space models: (a) contains one irregular anomaly, (b) contains three anomalies, (c) contains five anomalies.
  • Figure 2: Model construction process: (a) initial layered model, (b) model with added faults, (c) model with added folded strata, and (d) model with embedded anomalies.
  • Figure 3: Examples of OpenEM models, illustrating the nine types included in OpenEM.
  • Figure 4: Statistical information of OpenEM. (a) Resistivity distribution, (b) Number of layer distribution, (c) number of anomaly distribution.
  • Figure 5: Forward modeling network architecture, with the overall structure shown on the left and detailed module information on the right.
  • ...and 3 more figures